Anyone who’s ever done graph visualization has at one point or another run into the dreaded “Hairball” : Where the sheer number of connections overwhelms the data leading to a giant solid blot of connecting lines, obscuring anything useful. Robert Kosara has a short discussion of the problem and a few solutions to it such as PivotGraph and Parallel Sets.
For a while now, people in visualization have talked about the graph without the graph, i.e., graph visualization without the hairballs. Networks are clearly important and challenging data, and it seems a bit myopic to only look at node-link visualization. Node quilts and the PivotGraph represent promising steps into a very different direction. While they require more work to understand and are more limited in what they can be used for, they are also much directed towards a goal than just showing all of the data. I think that this kind of thinking will lead us to much more interesting techniques in the future than trying to teach the old node-link diagram new tricks.
